Look what the cat can’t drag in

Look what the cat can’t drag in


– So I mounted the DeepLens
camera above my cat door, and then I spent months sorting through and labeling over 23,000 images. His favorite time to do it
was three in the morning. Waking up in the night
to go to the bathroom, then discovering the corpse of a rat. People were like, “You
should just lock your cat in all night,” but he
poops exclusively outside. I barely have to clean his litter box. That’s a lot of talent to let go to waste. I decided that what I needed was, I needed that cat door to lock only when he is trying to
bring in dead or dying animals. I thought, let me talk to
my engineering friends. They pointed out that the tools for machine learning
used to be in the hands of only the most elite
computer scientists, and now somebody with a problem, like me, can go and figure it out. I thought, “I’m gonna try this.” The locking mechanism was the easy part. Basically just an electromagnet, and I attached to an Arduino. The hard part is how to get the lock to know when to lock. So I fed all my images up to SageMaker, and built three models. The first model just
asks, “Is there a cat?” The second model says,
“Is the cat leaving?” Or, if the cat is coming up the ramp, it calls its third state. If he is in killer mode, locks the cat door, second of all, texts me, and sends a pic, and third, donates blood
money to the Audubon Society. Looking back, it was never that hard, because of how accessible
these tools have become. But whenever my motivation
was running low, Metric would bring in a
fresh, horrifying new thing for me to encounter, and
got me right back there. So thank you, Metric, for that. You are dismissed, sweet beast.

Leave a Reply

Your email address will not be published. Required fields are marked *